Seeking State Recognition in Automated Policymaking: The Case of Clean Slate Criminal Record Expungement

• In Event: Automating Justice: State Policy Variations, Discretion, and the Impact of "Clean Slate" Record Relief

Abstract

Policymaking has increasingly embraced automation and algorithmic approaches to adjudicating benefits and determining eligibility. We use the case study of automated criminal record expungement—known as "Clean Slate" policy—where the state automatically seals eligible criminal records held in governmental databases without notifying the record subject, to understand how people impacted by this automated benefit understand the policy, view the role of the state, and interpret their changed legal status. Drawing on legal consciousness and legal ambiguity frameworks, we conducted an original survey (N = 750) and in-depth interviews (N = 30) with people eligible—but not necessarily aware of their own eligibility—for automatic expungement across five states (Colorado, Michigan, Minnesota, Oklahoma, and Virginia). We found that awareness of Clean Slate policies was low: while most respondents knew what expungement was, only 39 percent had heard of the automatic version that had been passed in their state. Upon learning about the policy, 94 percent believed the state should notify individuals once their records are sealed. 

Based on interviews and open-ended survey responses, respondents overwhelmingly valued official state documentation of their expungement, viewing it as essential "proof" of rehabilitation that could be leveraged in employment, housing, and other institutional settings where criminal histories are assessed. This desire reflected three underlying motivations: notification as a vehicle for public education about the law, as a source of confidence in navigating future opportunities, and as a form of data ownership over state-held personal information. Respondents also expressed skepticism that the state would fully deliver on the promise of a Clean Slate and preferred that information about the policy come from state sources rather than legal aid or community organizations. 

Our findings reveal a fundamental tension in algorithmic governance: while automation can efficiently deliver legal benefits at scale, the absence of recognition and documentation may render those benefits difficult to realize and leverage, placing the burden of proving state action back on individuals.

Authors

• Joanna Wahmhoff, Northeastern University
  Presenting Author

• Sarah Lageson, Northeastern University
  Non-Presenting Co-Author

• Ericka B Adams, San José State University
  Non-Presenting Co-Author

• Elsa Chen, Santa Clara University
  Non-Presenting Co-Author